Reinforcement Learning-Based Preset Trajectory Tracking Control for Vehicle Platoon Under State Constraints

Abstract

This paper investigates the preset trajectory tracking control problem of vehicular platoon systems (VPS) with state constraints and preassigned performance requirements using reinforcement learning (RL). Specifically, both velocity and acceleration constraints are simultaneously considered along with performance requirements, thus providing a more comprehensive guarantee of system safety and stability. Existing approaches based on barrier Lyapunov functions (BLF) can tackle these issues but are prone to nonlinear growth and singularity problems, which complicates controller design. To overcome these limitations, a novel two-step state transformation strategy is proposed. First, a nonlinear mapping function (NMF) is employed to reconstruct the states, embedding velocity and acceleration constraints directly into the transformed state space. Second, a preset-trajectory-based preassigned performance control (PPC) strategy is adopted to convert the performance requirements into a tracking task. This strategy effectively mitigates nonlinear growth and singularity issues, simplifying controller design and stability analysis. On this basis, a simplified RL algorithm based on neural networks (NNs) within the actor–critic framework is integrated with the backstepping method, to enhance the overall control performance. The proposed method guarantees internal stability and string stability of the platoon. Simulation results validate the effectiveness and advantages of the proposed method. Note to Practitioners—Vehicle platoon control is a key technology in intelligent transportation systems, playing an important role in improving traffic efficiency and ensuring driving safety. In practical scenarios, vehicle operations are not only constrained by velocity and acceleration limits but are also required to satisfy strict performance specifications. However, traditional BLF-based methods may lead to singularity issues and increased nonlinearities, thereby complicating controller design and reducing system reliability. To overcome these limitations, this paper proposes a RL-based preset trajectory tracking control strategy that employs a two-step state transformation. The proposed approach not only guarantees system safety and prescribed performance but also effectively enhances control performance. Overall, it provides a practical and safe solution for vehicle platoon control in complex environments. Practitioners should note that parameter tuning is necessary to adapt the framework to different traffic conditions, and future implementations should further account for obstacle avoidance.